Robot Density and a New Job Loss Forecast

Introduction

Over the last few years, researchers have published a number of studies about the impact automation will have on the job market. However, the results have ranged wildly, pegging potential job displacement in the United States at anywhere between 9% to 47%.Surprisingly, though, these studies all tend to ignore the material basis for automation: The actual production and penetration of robots.

We figured we needed to develop a new methodology, and we’ll explain that here today. We’ll analyze available cross-country data on robot density and labor productivity and propose our own alternative estimate for potential job loss. We’ll compare our results with other recent studies.

Obviously, many factors will determine the scale of job loss, but a simple ratio — robot density to labor productivity — offers a clean and accurate forecast. It also gives us a good idea of how job displacement will be distributed among countries.

Robot Supply and Density

Data shows that since 2010, the trend toward automation worldwide has considerably accelerated the demand for industrial robots. For example, between 2012 and 2017, robot sales demonstrated the strongest growth annual rates on record, and between 2011 and 2017 the average annual supply more than doubled compared to the average between 2007 and 2010.2

The main drivers of the exceptional growth we see between 2015-2017 were in electronics (+86%) and the metal industry (+55%). The automotive industry remains the largest single-industry consumer of industrial robots — sales increased by 29%, and the industry claimed a 33% share of the total supply in 2017.2

Five geographic markets represented 73% of global industrial robot sales in 2017: China, Japan, Korea, the U.S., and Germany. China’s steady economic and manufacturing growth has made it the dominant market since 2013, and it’s been expanding. In 2017, about 137,900 industrial robots were sold to China — up 59% from 2016 — and the country was home to more than a third of the total global supply. However, the market share provided by Chinese robot suppliers was only 25%. This means China acquires 75% of its robots — including ones assembled domestically —  from foreign sources.

Density, Productivity, and Job Loss Potential

Since 2012, the global stock of operational industrial robots has increased considerably, as has the growth rate.2 But sales volume and stock don’t accurately reflect the size of the economy. Our hypothesis turns on a different metric: Robot density, which the International Federation of Robotics defines as the number of multipurpose industrial robots in operation per 10,000 persons employed in manufacturing.

Here’s that hypothesis — it’s pretty straightforward: The higher the robot density, the higher the labor productivity. We can use the relationship between those indicators to estimate the number of jobs at a given level of density. Here’s what we found.

In 2017, average robot density was 85 robots per 10,000 employees globally, but it wasn’t evenly distributed. The countries with the highest densities — by far — are the Republic of Korea (710 per 10,000 employees) and Singapore (658).2 The gap is remarkable. Germany and Japan, for instance, have about half that density. Also worth noting that from 2009 to 2017 China’s density increased nearly ten-fold.2

We collected data from several recognized sources:

We drew on data from 43 countries, and adjusted the value-added in manufacturing to account for differences in price levels. And to account for short-term fluctuations in labor productivity — as well as to account for any observational variations due to missing data — we used the five-year productivity average between 2012 and 2016.

For those of you interested in the wonkier details of our methodology, we estimated the relationship between productivity and density based on a simple OLS equation:

LN Lprodt=α+β*LN Rdenst+e

Where

  • LN Lprodt is the natural logarithm of the 2012-2016 simple average of value-added per person employed in manufacturing at PPP dollars;
  • LN Rdenst is the natural logarithm of robot density; 
  • coefficient represents the percent change in productivity for every percent change in density;
  • α is the constant; and
  • e is the error of approximation.

Table 1 summarizes our estimates:

Table 1. Estimates of Labor Productivity Equation 

Annual data 2016
43 countries
β0.276
(0.038)
α3.121
(0.160)
Obs43
Adjusted R20.57
SE0.34

Note: Standard errors are in parentheses.

Our model suggests that a one-percent increase in robot density corresponds to a 0.3% growth in labor productivity. We can then use this ratio to estimate the share of jobs susceptible to automation. For our density benchmark, we took the last available 2017 value for South Korea, which, as noted above, is 710 industrial robots per 10,000 manufacturing employees.2

Assuming robot density in the United States eventually increases from 200 to 710 — and given 2016 technology levels — we project nearly one in three (29.7%) U.S. manufacturing jobs is currently at risk from automation. In China, about half (47.6%) are susceptible, and Russia, the Philippines, and India could see displacement levels of almost 80%.

Excluding current state of robotization from analysis leads to contradictory results

We compared our results to two other studies,3 4 and found that when you account for the current state of robotization, some job loss predictions might look contradictory.

The U.S. and Russia are just two examples. According to our estimates, based on the difference in levels of robotization, manufacturing job loss potential in those two countries are 30% and 78%, respectively, with respective robot densities of 200 and three per 10,000 employees. This seems intuitive: The fewer robots in use today, the greater the potential for future job displacement.

In contrast, studies by PwC and the McKinsey Global Institute suggest job loss potential is actually higher in the U.S. than in Russia, despite the fact current robot density in the U.S. is 67 times higher than in Russia. These studies weigh mitigating factors — such as lagging technology — which we discussed in greater depth in an earlier article in this series. (Take a look back at Measuring Technological Impact on Employment.)

Conclusion

Robots will displace jobs globally, but that displacement among different countries will happen at different levels and different rates. It’s a complicated calculation, but we believe a simple ratio of density to labor productivity offers an alternative methodology, and we’ll follow up on it as fresh data becomes available.

References

  1. Measuring Technological Impact on Employment. Knoema, 2018. Link
  2. World Robotics 2018 Industrial Robots Executive Summary. International Federation of Robotics, 2018. Link
  3. Where machines could replace humans — and where they cant (yet)”. McKinsey Global Institute, 2016. Link
  4. John Hawksworth, Richard Berriman and Saloni Goel, 2018. “Will Robots Really Steal Our Jobs”. PwC. Link

Making Sense of Global Demographic Megatrends

Introduction Over the last few years, researchers have published a number of studies about the impact automation will have on the job market. However, the results have ranged wildly, pegging potential job displacement in the United States at anywhere between 9% to 47%.1 Surprisingly, though, these studies all tend to ignore the material basis for automation: …

The Future of Global Conflicts

Introduction Over the last few years, researchers have published a number of studies about the impact automation will have on the job market. However, the results have ranged wildly, pegging potential job displacement in the United States at anywhere between 9% to 47%.1 Surprisingly, though, these studies all tend to ignore the material basis for automation: …